Originally Posted by notnyt
Well, Anechoic/Herb calibrated this EMM-6, he can chime in with how close his calibrations are if he chooses. Judging by the resolution used and the procedure, I don't assume it would be very far off.
My cals (probably) are not +/- 0.5 dB accurate, certainly not across the full bandwidth of the measurement. If we start with the reference mic (which I get regularly calibrated by Scantek
, an NVLAP/NIST-traceable lab), their cal methods have an uncertainly of 0.2 dB from 200-8000 Hz, 0.6 dB from 8-10 kHz and 0.9 dB from 10-20 kHz using an electrostatic actuator.
I haven't done a complete statistical analysis of the accuracy and precision of my method, but I've compared results I get with my setup with the results Scantek has generated for my other ANSI-rated mics to get an idea of the accuracy.
I use a free-field substitution method for the data about 100 Hz. The results about about 8 kHz is extremely sensitive to mic placement and room effects which cause most of the error. Below 100 Hz I use the substitution method in a pressure chamber and the error comes from imperfect seals around the mic capsule in the coupler (this is an issue with the EMM-6 and ECM8000 since they are slightly smaller in diameter than my 1/2 ACO 7052, so I have to pad out the diameter with electrical tape to get a good fit in the coupler).
Given the above, the data I provide from 100 Hz to around 3-5 kHz is accurate to around +/- 1 to 1.5 dB. Below 100 Hz and above 8 kHz or so the error increases, probably in the neighborhood of +/- 2-3 dB (perhaps more in the highest or lowest octave). The substitution method in principle is relatively simple but in practice there is a lot that can cause error. I think this is one of the reasons why NIST-traceable labs don't bother with this method (they all use electrostatic actuators now).
(in another thread either here, on hometheatershack.com htguide.com, or parts-express.com I have some accuracy numbers that were based on some quick calcs I did on data, and this post is an attempt to remember those numbers. If there is any conflict between this post and the earlier post, go with the numbers in the other post).
I do know that my method is repeatable. Every once in a while, I have an overseas delivery go awry and I get a microphone returned to me after it's been sitting in a foreign post office for a month or so. When I test it again, the result is always within +/- 0.5 dB of what I got previously. Thank heaven for small favors.